Are My AdWords Paid Search Ads Performing Well?

Are My AdWords Paid Search Ads Performing Well?

Kevin Dieny

FIRM FIXED IDEAS

Adwords Paid Search
Google Ads Quality Scores
Ad Relevance

How to know if your Google AdWords paid search ads are performing well and what to do about campaigns that are underperforming.

“Well paid search marketers spend their time trying to eliminate their errors. Great paid search marketers spend their time looking for new opportunities.” – Me

To ensure that your ads for AdWords paid searches perform well, you first need to define what “well” means for your AdWords account. To do this, you will have to ask yourself two things:

  1. What are the right goals for each particular campaign?
  2. How and where is success measured?

Before jumping in, I’m going to assume you know enough to be dangerous in AdWords and have the fundamentals down. If you want to learn the fundamentals or want a refresher course on Google AdWords or Paid Search, then I recommend you first jump over to Lynda.com and take the Essentials Course.

Why didn’t I recommend you take the Google AdWords Certification Course + Test? Simply put – I am not a fan. I do not think it provides the strategic mindset that is most beneficial for marketers who run AdWords accounts. For that, I recommend taking the Lynda course mentioned above or a course that gives you strategies and proper expectations for running campaigns.

Now, let’s get into it.

Launching a paid search campaign with adequate expectations

You cannot know if your AdWords campaigns are successful unless you have given them appropriate time and you have set them up for future optimization (that you plan to do). I agree with Brad Batesole, the author of the Lynda course, when he says, “A typical AdWords campaign will hit its stride in about three months.” It can be painful to look at your dashboard every day for hours on end and stare at that 0 conversion number. You need a better hobby….

Let’s start with the crucial elements you have to get right for any Adword campaign to be successful.

  1. Ad Position – where on the Google search page your advertisements will appear, in a numbered and ordered list, where position 1 is the most desirable and 10+ is the least desirable position for your ad. (Update: Consider Impression Share if Ad Position has deprecated).

2. Relevance – (aka Ad Score) a mixture of the quality score based on the congruence of keywords being matched, advertisement copy being served, the landing page performance metrics, and the history of success.

3. Auction-Based Bidding – bidding on keywords based on the demand for each keyword at any given time, in order to be the highest bidder and win your position on the search page, so your ad is served to the searcher.

What’s the problem then? What is unclear about current multi-channel attribution?

Bringing clarity to multi-channel attribution tracking is best illustrated by the models that attempt to describe the fluid dynamics of the process. Before we get into the models, let’s focus on the concept.

Consumers will digest or consume lots of content/information from the channels they prefer.

For example, when you launch an Instagram only campaign, you’ll miss your targeted audience who doesn’t have Instagram. Sounds obvious, right?

Small and medium sized businesses assume their customers are only in a few channels. However, without good data, you don’t know what you don’t know. You would need to hit as many channels as possible to correctly determine which channels your customers are actually consuming information from.

In order to clear up the attribution problem inherent in multi-channel marketing, you need to first understand the role of your channels as you scale. Once you do this, you can see why accurate attribution is a struggle. Unless you can attribute channels correctly, you will not know the true impact of that channel on ROI.

Your consumers are more than likely in all of the channels.

Just based alone on the above image, this seems like a lot of marketing channels—and it is. But truth be told, there are many more. Anywhere a consumer can interact with your business in a tangible or digital way – that’s a channel you need to be in.

(Tip!) If you’re feeling overwhelmed, a great place to start is by looking to your competitors to see what channels they are in.

Start with the goal in mind

First, can you go down your list of campaigns and tell me what the goal is? Yes—one singular, overarching goal. Ad Groups should be distinctly aligned to this same goal, too. Brad also suggests that you theme your campaigns – I am a fan of this, as I even distinctly theme my own ad groups to the goal and match types I am utilizing.

Once you have your goals broken out, can you identify the KPI (key performance indicator) that you will measure for them? Here is a list of goals and their typical corresponding KPI’s; remember to make yours as specific as possible:

GOAL KPI (must be specific)
Awareness of my brand Impressions
Traffic to my website Unique Clicks (true click-through*)
Phone calls to my business Clicks to call + Tracking number calls
Customers to walk into my business Impressions, Clicks/Calls to directions
Sales Conversions, offline-conversions

*a true click-through is when a click occurs along with a corresponding page view

Write yours down.

Great!

Now, here’s a great question to ask every time you are scoping out a new campaign. Or, you can even apply it to existing campaigns to see if they follow this rule. Who is the best audience for this goal?

First, we may need to define what am I talking about when I say audience.

Building your customer avatar or persona that aligns with your intended goal.

The best way I have ever seen a customer avatar built is from a worksheet provided by Digital Marketer. Head over there and grab their sheet because it is an amazing foundation for getting started on persona work.

Once you have your avatar built, make sure you have checked off each of steps we’ve outlined so far. It is a best practice to go step by step through your AdWords account to make sure you are not committing egregious errors and are allowing your campaigns to be successful. By eliminating your errors, you can turn your focus to more creative prospects for your account.

Make sure you can answer these questions:

  • What motivates your ideal customers? Be specific.
  • Why should your ideal customers care about your products/services? Be specific.
  • What is the specific pain that you want to solve for? Be specific.

Noticing a trend here?

At the heart of knowing if your AdWords ads are working is this fundamental truth of marketing – the best marketing is specific and personalized. Nine out of ten errors in AdWords campaigns are committed by being not specific enough or aligned well with your ideal customer. You have to learn your customers’ pains, their identities, and ultimately how they search online (what keywords or slang they use).

What does this kind of specific look like?

Well, imagine that you said my ideal customer is 30 to 45 years old, usually a female, and works full-time. That’s broad – you can do better. So let’s enhance that.

Now you say well… actually, our best customers are usually 40 to 45-year-old females who are married to white-collar partners. Both work full-time, and they need to live within 45 minutes of my stores. That’s better, but you can still get more specific!

Always start as precise as possible and adjust back the specificity if you need more impressions or are finding you have missed the mark. So let’s again enhance what we already have in this example by adding that our ideal customers are those families who make more than $50,000 a year, use their phones primarily for searching, rely on reviews, are price sensitive, and browse the internet while at work and call during their lunchtime. Bam! That is a lot better. Of course, you could always go deeper, but this works. Here are some categories to get you started:

  • Age
  • Gender
  • Marital Status
  • Education Level
  • Primary/Secondary Language
  • Location
  • Job Title
  • Income
  • Industry
  • Family Makeup
  • Technology Usage
  • Web Access
  • And so much more…

Eliminating Errors and Refining

I’ve already touched on how it’s an acceptable norm to give your AdWords campaigns at least three months (90 days) to adjust. If you skimp on this, you will likely always see poor results and give up completely on trying to get sustainable wins in AdWords.

But I want to get more detailed now that we have the basics out of the way. The crux of the problem isn’t that you need to be the best AdWords marketer ever. Instead, it is simply to find out if your ads are working well. “Well” in this context means that you’ve got your kinks and errors worked out and now you are just trying to refine.

Here is my list of refinements in the order that I would assess your account:

  1. Am I familiar enough with the AdWords interface to know where to go to find my KPI metrics that I have planned for my goals?
  2. Do I have a report with my KPI metrics along with the corresponding campaigns/ad groups/ads and keywords for the desired specific goals?
  3. What are my quality scores for my ads?
  4. Do my ads need better relevance?
  5. Have I set up and optimized my landing pages for the advertisements?
  6. Have I set up and tested my conversion tracking to ensure that I am receiving conversions into my AdWords account?

1) Familiarity with AdWords

To get more familiar with the platform such as where everything is located, where to go to answer your question, etc., you simply have to spend time in there. Google has a plethora of guides and searchable help to lead you where you need to go. AdWords has the capability of solving more questions. If you are still stumped, there is also the reliable online search to perform.

2) Report Building

AdWords has a new(ish) reporting feature built right in that you can use to search for and build reports based on your own campaigns. I caution you that it does take some getting used to, so start playing around with your reports. Look online for examples of reports others have made for good examples. Try not to overdo your reports – stick with your KPIs.

3) Quality Scores

Quality scores are attached to keywords, but they also correspond to your ad copy and the destination landing page. Quality score is Google’s measure of how relevant the experience of the user is to the ad. The higher the score, the less you pay. This is set up this way because Google wants to give users the best advertisement experience they can. The more people spend on ads, the more money they make – and they will only do so if it’s a good environment to do.

4) Relevance

Try to make the experience of the person clicking on your ad seamless from start to finish. You want the colors and the wording to be as similar as possible. Manage their expectations. If you tell them they will be learning more, then teach them! Do not ram a lead magnet down their throat. If you say, “see the difference?” then show them the difference. It’s shocking how much these tenets are violated.

5) Landing Page Optimization

Starting at the end, go to your landing page and see what keywords are the most common on the page. Do you have a privacy policy and/or a terms and conditions link available to visitors on that page? Do you have your contact information located on that page?

You also need to make sure the page loads quickly, does not have damaging pop-ups and that the copy of the page aligns with your keywords and your ad copy. There is a lot that can go into landing page optimization, but I will save that for another article.

6) Conversion Tracking

You have two places you need to go to ensure that you have setup your conversion tracking. First, your tracking code (tag manager) needs to deploy the tracking event. Second, your AdWords account should have a conversion setup. Go to the conversion menu inside AdWords and make sure this is set up with a value. To finish, test it and make sure it fires and all the data is recorded accurately. Your AdWords will be lightyears more accurate if you are tracking your conversions into AdWords.

Now what, I’ve done everything!

If you can check off each list, I’ve delivered and are still having issues of knowing whether or not your ads are working well, then consider these big-picture metrics. Aim for the best but always try to make fine adjustments over months of time (things won’t work better overnight most of the time unless you goofed). Remember, what truly matters is making your goal happen.

Ask yourself if you have the below industry standard metrics. If your answers are “yes,” you are probably doing well and have enough information for a decision of whether or not to change something:

  • A click-through rate above 2%?
  • An average position at or below 1.9? (1.0 is killing it!)
  • Keyword quality scores at or above 7?
  • Impressions above 1000?

Notice I didn’t say anything about the cost per click, conversion rate, view-through conversion rates, match types, search terms, or ad extensions. This is because this is very subjective to your individual campaign. A guide that delves into these has to consider that if you are using them, it aligns with your goal. Congruence with your goal is key. I would recommend evaluating your search terms report and taking advantage of your ad extensions… but how you go about maximizing this depends on your campaign goal.

You need always to be striving to find new customers. But showing the right relevant message and keeping your campaigns specific takes patience; optimization takes time. Learning to be patient is the only pain I cannot give any advice on. It’s not easy.

Thanks, I hope that now you can say, “Now I know I need to define what well means for my campaigns before I can see if they are performing,” and “I know how to find out what success is for my AdWords account.” If you have questions, comments, or ideas, feel free to add them to the community to see in the comments section below.

Kevin Profile Shot
Kevin Dieny
Marketing Professional

Kevin is a contributing author and product expert in all things digital where he works as a Digital Marketing Analyst at CallSource.

Bringing Clarity to Multi-Channel Attribution Tracking

Bringing Clarity to Multi-Channel Attribution Tracking

Kevin Dieny

FIRM FIXED IDEAS

Attribution Models
Multi-Channel
Tracking

Businesses who understand their complete customer journey can capitalize on opportunities and unlock the full set of experiences which lead to more sales.

Let’s break out the chalk and chalkboard!

​Businesses who understand their complete customer journey can capitalize on opportunities and unlock the full sets of experiences lead to sales.

There are two basic ways in which we can break down the complete customer journey, and we can use either of these to model how we want to attribute channels across the marketing spectrum. Measuring true ROI and the impact of our marketing efforts on the business depends on our attribution model.

But hold up…I am getting ahead of myself. What is the customer journey and what are channels, attribution, and the marketing spectrum, you ask? Let’s break out the chalk and chalkboard!

Here are the basics:

Channels are targeted sources that produce traffic for your brand, website, company, store, etc. You can think of channels as places where people find out about your products and services. Advertisements and social media are the most common channels right now. Because these types of channels are easily measurable, they are insightful. By understanding what channels your consumers are interested in, and adding your content to those targeted sources, you will be able to send specific traffic to your brand.

Attribution is being able to identify the source, or channel, of traffic coming to you. If a potential customer walks into your business because they read about it in the newspaper, then the newspaper is responsible for sending them to you—that is the channel that you can attribute to this new customer.

The age old problem that can occur with determining accurate attribution is figuring out which advertising source/channel deserves credit for sending a customer to your business when that person may have read about you in the newspaper and also saw a billboard advertisement. We will delve further into this issue a bit later.

The marketing spectrum is the full range of activities that you are performing to engage with customers in order to build value for your brand, sell products, and deepen relationships. Activities may include emailing, calling, texting, advertising, social posts, and sending direct mail. Since there are numerous ways of interacting with customers, it is important to consider all the different channels and activities in those channels that are contributing to building your business.

What’s the problem then? What is unclear about current multi-channel attribution?

Bringing clarity to multi-channel attribution tracking is best illustrated by the models that attempt to describe the fluid dynamics of the process. Before we get into the models, let’s focus on the concept.

Consumers will digest or consume lots of content/information from the channels they prefer.

For example, when you launch an Instagram only campaign, you’ll miss your targeted audience who doesn’t have Instagram. Sounds obvious, right?

Small and medium sized businesses assume their customers are only in a few channels. However, without good data, you don’t know what you don’t know. You would need to hit as many channels as possible to correctly determine which channels your customers are actually consuming information from.

In order to clear up the attribution problem inherent in multi-channel marketing, you need to first understand the role of your channels as you scale. Once you do this, you can see why accurate attribution is a struggle. Unless you can attribute channels correctly, you will not know the true impact of that channel on ROI.

Your consumers are more than likely in all of the channels.

Just based alone on the above image, this seems like a lot of marketing channels—and it is. But truth be told, there are many more. Anywhere a consumer can interact with your business in a tangible or digital way – that’s a channel you need to be in.

(Tip!) If you’re feeling overwhelmed, a great place to start is by looking to your competitors to see what channels they are in.

Where to get started with multi-channel marketing?

Until you know where your best customers are primarily found, you should focus on the largest and most universal channels first. Depending on your business and its goals – consider the following channels to start marketing into:

  • Website (CMS) – You need a digital location no matter what type of business you are; this is foremost the most important thing.
  • Search Engines (SEO) – You create a presence in search by creating quality and relevant content that corresponds with the products you sell and what your ideal consumers want to know.
  • Social Media/Listing (Distribution) – You need to get your content to people who need it. For people looking for your services or products that you offer, you must have a social presence in order for them to find you on listings or review sites.
  • Ads, Email, and Sales (The Funnel) – This is the final step that allows you to build your marketing “machine.” The machine will help you drive traffic to your content, convert consumers at the end of their journey into leads, and eventually move them down the funnel from a lead to a sale.

Once you have the primary channels out of the way, you can begin inserting your business and managing your presence across the entire spectrum. This exercise should help you see the role of each channel for your individual business.

A word of warning: scaling into more channels requires more content, which always requires more time. As you widen your breadth of channels and learn more and more about your ideal customers, stay mindful of the time constraints this presents to your business.

The attribution models of a multi-channel spectrum of marketing activities:

There are numerous models that attempt to attribute the deserved value to the proper channel. While each model has its pros and cons, I first want to cover the most popular model, “Last Click.”

Last Click is the most popular because even if your cookies or tracking is unreliable, this will always deliver reliable results based on a consistent method. Google has their own write up comparing models present in Google Analytics you can utilize in reporting.

Last Click/Final Click Attribution Modeling

Last Click attribution is attributing the final channel’s click through to the source of lead generation. Assuming several channels (which could be near infinite) had touches to the consumer along their complete journey, Last Click will attribute the channel that they touched last to receive the credit in converting that lead.

Example #1 – A consumer sees an advertisement on television, and then sees that advertisement again later on YouTube. Later, they get an email promoting the advertisement, and finally they decide to Google the company before buying. In this example, Organic search gets the credit and the attribution for converting that person into a lead.

PROS CONS
Most reliable for cookie based tracking. Negates contribution of prior channels.
Simplest to use; usually the default model. Doesn’t usually reflect marketing efforts.

First Click Attribution Modeling

First Click attribution is the reverse of Last Click. The first channel that brought someone into all of the other channels is given the credit in lead generation. Assuming again that there were several channels (this could be near infinite) that had touches to the consumer along their complete journey, First Click gives attribution value to the very first channel we had tracking information on.

As I mentioned, most attribution models prefer Last Click because of the unreliability in tracking. First Click attribution is the most susceptible to unreliable tracking. For example, if someone deletes their cookies from a cookie-based tracking system, then you get unreliable first clicks, as their first channel will be reassigned after their cookies were purged.

Example #2 – Imagine that a consumer sees an advertisement first on television, then watches the advertisement on YouTube. They then receive an email promotion for that advertisement, and finally they search and convert to a lead from organic search before buying. In this example, the very first click or channel, namely television, will get the credit for the attribution.

PROS CONS
Emphasizes awareness strategies. Most unreliable cookie based tracking model.
Used in brand awareness and political campaigns. Entry channels over-weighted.

Linear Attribution Modeling

Linear Attribution modeling was statisticians’ go-to before exploring years and extended data. In statistics, it is a rule of thumb to first assign all weights a consistent value before attempting to individualize the weighting of every element. Based on this assumption, I would assume this is where linear became a popular marketing model.

The purpose of assigning equal value along the chain is that it may be too complex and we may not actually know enough to weigh any specific channel greater value than another. This model excels at awareness and recognition strategies, but does not help you do much else. Linear Attribution also has the flaw that if something did have a greater impact on the whole marketing spectrum, we would not be able to identify it correctly. But keep in mind—there are marketers that swear by each and every model, so do your research to determine which model best suits your business.

Example #3 – Assuming the same story as the examples prior, each click along the complete customer journey is assigned equal value as long as the consumer eventually converts.

PROS CONS
Simplest (think statistician). Nothing is given uneven weights.
Does not emphasize any particular strategy. Cookie cutter marketing strategies.

Time Decay Attribution Modeling

Time Decay attribution is a confusing model as you need to know at which point the greater credit is assigned to a channel. In Time Decay modeling, the channels at the end of the consumer journey receive the most credit and the channels at the beginning receive the least. The channel at 50% of the consumer journey would therefore receive 50% of the weight.

Time Decay is helpful if you see a lot of your channel traffic from season and clustered events. You can weigh conversions nearest to the channels that led to a conversion more and give less credit to channels that contributed in a time period that may not have contributed to the consumer’s final decision to convert. Again, decide what model best reflects how your business does marketing and match the model to the type of improvements you want to focus on.

Example #4 – Assuming the same story as the examples prior, each click along the complete customer journey is assigned increasing value as long as the consumer eventually converts. Another way to describe it is that each click’s value decays over time as more channels and clicks are added.

PROS CONS
Gives credit to clusters of clicks. Assumes that prior clicks are worth less.
Helpful if you have most of the market share already. Favors a bottom of funnel centric strategy.

U-Shaped or Position Based Attribution Modeling

Position Based attribution modeling is one of the models that actually favors a certain type of marketing strategy. Position Based attribution favors straight lead generation as it gives credit to the beginning channel that brought someone in and the final channel that converted. This is why it has a U shape, as seen in the diagram below.

The position based tracking does suffer from cookie based tracking, as does any model that relies at all upon reliability. However, this model does have strengths in giving credit to marketing efforts at the ends, where the big impact actions happen. Giving credit to the end and beginning gives credit and therefore rewards the acquisition channels of marketing; which is why it lends itself to lead generation strategies.

Example #5 – Assuming the same story as the examples prior, each click along the complete customer journey is assigned a value, but when the conversion happens, the first click and the last click are given much more credit.

PROS CONS
Focuses on two important clicks. Doesn’t help you build complex strategies.
Leans towards lead generation strategies. Ignores values of relationship building.

Special Mentions of Attribution Modeling

I could go on with the other 7 or so advanced attribution models and another dozen or more of the super advanced or specialized attribution models, graph them, and weigh their pros and cons, but I’d rather touch on how all of these are implemented and tracked, as that is the intent of this article.

I will instead list the others and attempt to cover the deeper look into all models at a later date. Here are some of the more advanced iterations of attribution models that exist alongside the fundamental attribution types mentioned already:

  • Middle of Funnel or W-Shaped Attribution Modeling – This model brings the valuation of channels that extend into channel actions at the middle of the funnel but not up to the point of a completed sale.
  • First or Last Touch/Email/Conversion Based Attribution Modeling – Here, a specific type of action with a channel identifies as the first or last attribution; giving it full credit for the role in creating that touch.
  • First or Last AdWords Attribution Modeling – Google points out this type of attribution model in its description of analytics models you have to choose and report on. The benefit here is that you give credit to search marketing efforts solely – in order to calibrate paid search and make it more efficient but it somewhat ignores the other channels. To me, this modeling is more ad-hoc.
  • Full Funnel or Z-Shaped Attribution – Full Funnel is similar to linear modeling, but the model extends beyond a conversion or initial sale to the final sale and weighs all of the channels that have an impact along the way with equal credit. This is the most widespread channel attributor but it does not specifically weigh any one place in the funnel as more important than the rest – a concept that bottom or top of funnel marketers scoff at.
  • Attribution Modeling in R – Analyzecore has a great article discussing the practical concepts of modeling in R code or algorithmic systems based modeling, the point of which is to conceptualize the flow and customize it to any model that is desired on the fly. This and other models that rely upon more complex algorithms are expandable if you have the desire and need to dig into your data on a highly granular level.

The problem with attribution modeling is unreliable tracking.

Hinted at throughout this article is that the problem with attribution modeling comes down to reliable tracking systems. As a consumer progresses along their multi-channel journey they should be tracked so that every touch with a channel is measured to see, at minimum, that it happened. But there are some struggles with even getting that much information because of the way we track that kind of interaction.

Measuring: The Silo of Marketing Stacks

As marketing teams add new tools and measurement analytics to their stack, they end up adding a silo of information that is separate from all of their other tools. In order to merge and connect that information, it requires setting up integration or API connections so that the technologies in your stack share and consolidate information. This is great when all of your stack tools cooperate or a nightmare if you are stuck exporting into CSV or XLS files and swapping that information around.

Assigning: The Unique Identifiers that Link Consumers to their Data

All marketing tools assign a unique identifier to every single point of data that is connected back to the source of that connection, typically a consumer. Consumers are often listed in marketing tools with all the data that corresponds to them (tools vary), but there is always a way for their systems to deal with duplicates. This is usually done with unique identifiers, or sometimes it is done with emails as these are the most basic unique identifier in the industry. Other systems provide long hex codes with near impossibility of duplication in order to ensure that every single person is accounted for (and it actually works).

Tracking: Building Consumer Databases with Input

Databases in tech silos are assigned unique identifiers after they are inputted into those systems. The technology that adds them to these systems are processes that often rely upon tracking (at least in the analytics world they do). In order to track consumers, we need something robust enough to follow them on properties that we control. The most common analytics tracking code in the world is Google Analytics, installed with a simple few lines of Javascript code into the header of webpages. In my opinion, 90% of all tracking processes are based upon cookies.

Since they have been referred to multiple times in this article, let me simply clarify what cookies are. Cookies are miniature databases stored in the web browser of users which allow us to literally “follow the user.” Non-cookie based tracking does an end around the storage of information in the user’s browser and skips all the pitfalls of cookie based tracking, but requires elaborate privacy controls. For more information on cookie vs. non-cookie based tracking, check out my previous article here.

There are always pros and cons, but the point I want to make is that (for analytics) consumer information is tracked into marketing stack tools and linked to identifiers based on the robustness of tracking.

Clarity: Putting it all together

Gathering the information is the first step, and because of that, makes all future steps dependent upon the quality and robustness that comes from tracking and inputting that data.

This is why it is crucial to have reliable tracking as your ability to assign correctly is based on having enough information to make the consumers in the database distinct from each other. Further, it is vital that we measure; otherwise we cannot improve – hands down. Measuring is dependent upon having properly assigned consumers in our database, and upon robust tracking inputs.

Having any data is great. Even 50% accuracy of data gets us 50% of the way to 50% right and wrong decisions. Having no data and making decisions is called “Empirical Decision Making” and it’s best summarized as shooting from the hip.

(Tip!) Avoid it at all costs; do not shoot from the hip when it comes to making marketing decisions!

So the big question looms:

How do you attribute all of your marketing (which relies upon measurement) from the consumers (which rely on assignment) back to the channels that triggered those interactions (which relies on tracking)?

ANSWER:

You do your best. If you are trying to clear this issue up because you already are familiar with this issue, then you need to find a better solution that includes:

  • Tracking
  • Assignment
  • Measurement

You cannot even being to model your attribution and give credit to channels unless you understand the reliability of your entire spectrum.

Recommendations

Identify holes that you have, so that you can see where there is drop off in channels and assess whether that is a content failure or simply a tracking issue. Utilization of UTM standards and standardized naming conventions to go along with them is key. Cross-train marketing members on your team to help them understand the value in tracking as it applies to measurement and goal attainment.

Kevin Profile Shot
Kevin Dieny
Marketing Professional

Kevin is a contributing author and product expert in all things digital where he works as a Digital Marketing Analyst at CallSource.

How Do I Track My Customers on My Competitor’s Website?

How Do I Track My Customers on My Competitor’s Website?

Kevin Dieny

FIRM FIXED IDEAS

Cross-Domain Tracking
Google Tag Manager
Data Quality

You can deploy certain tricks to track users across domains, devices, browsers, and time but there are limitations in tracking codes.

The fundamental problem with competitor tracking is ownership.

Tracking customers on your competitor’s website.

You can deploy certain tricks to track users across domains, devices, browsers, and time but there are limitations in tracking codes. Since you are limited to what you are currently using, let’s attempt to solve the problem of tracking consumers onto your competitor’s website with what you are already using.

The fundamental problem with competitor tracking is ownership.

Since all the tracking codes embedded on websites all over the web are running all the time, you’d think that there would be a way to find out what happened after consumers leave your website. The short answer, sorry… you don’t get to see it. Why? You see, the reason you do not get to know is that you may only see data from tracking codes that you own and utilize. It would require you to have your tracking code installed on every website to get all of that tracking data.

Frustrating, right?

Well, then how do websites obtain competitor data? At least for now, as of writing this, everyone’s information is considered private. The web browsing data across the internet is tracked by dozens of big data companies who have access to every businesses’ competition, but that information is protected by privacy laws.

For the most part, an individual’s data is private, secure, and encrypted. When large data queries are made to analyze the data and pull out insights into competitor’s, every user is made as anonymous as footprints in the snow. This means that marketing research and analytics tools can have access to trackable data, but at a steep price for those with deep pockets and access.

 

Here are the basic steps you can follow to widen your tracking capability:

  1. Switch to a Tag or Code or Pixel Manager
  2. Utilize Cross Domain Tracking
  3. Improve The Quality of Data with Custom Dimensions

Step 1: Switch to a tag management solution

The most common manager is Google’s Tag Manager. I will refer to this as that is what I have been using for years and it allows me to make manageable changes and customizations to a vast amount of websites.

I would recommend that you take the time to set it up with the best practices of implementation including conventions, standards, protocols, etc. The result will be a single-stop-shop of all things tracking.

Step 2: Utilize Cross-Domain Tracking

I will again reference this regarding Google Tag Manager because that is what I have been using and what I know best. Before you can start utilizing Cross Domain information you need to set up your tracking code to allow for it. You’d want Cross Domain tracking if you want to track individual users that hop around on your websites. The alternative is that every visit will be seen as a unique user and your traffic numbers will be inflated with artifacts.

 

Within Google Tag Manager, using the newer interface of creating a Google Analytics Settings Variable, you will see the Tracking ID field, and the Cookie Domain field, at the beginning of creation. I am skipping beginner steps here to focus on the ability to implement cross-domain tracking between websites you own and have tracking codes on.

The setup for one of our variables may look familiar. The value you want to consider is the cookie domain field, entered as “auto.” Second, you will drop down More Settings, and Fields to Set to allow inter-linking as seen above. In the field value, type “allowLinker” or select it from the list that will appear. Enter “true” for the value. Now you are ready to add your referral websites. I will give you a shortcut after I explain.

Referral lists are the URLs of the websites that you own with tracking codes that will be sending traffic between them. This is done to facilitate the cross-domain transfer of individual users between the sites you list. So gather up the list of your websites and write them down. We are going to add them to a constant in Google Tag Manager.

The final field in the Google Analytics Variable is the Auto Link Domains field. Here you will click the grey “add” block on the right and add a constant. In the constant value add all of your website domains and subdomains one after another, separated by commas. Example: “callsource.com, marketing.callsource.com.”

Taking control of your Cross Domain tracking ends inside your Google Analytics. Navigate to your Google Analytics, and within each website, you need to add referrals to all the others.

This sounds tedious. It is.

There are workarounds, but this is the manual approach.

The referral exclusion list is located in your administrative panel, under the property, and finally nested under Tracking Info. I feel for you if you have over a hundred websites, I really do. For most of us, there might only be one to twenty websites to add and keep track of.

Once you have added all of your websites here that could be sending traffic, you are done. You should be tracking multiple sites now and that traffic will not fall under referrals and will be associated across domains that you own. It’s a work of beauty.

Step 3: Improving the Quality of Tracking Data with Custom Dimensions

This final step is a deep-dive, and if it’s worth the effort for you to learn this, I would ask why you haven’t considered trying out Digital AI. I need to add that this last section contains information that I am sourcing from a heavy hitter in the analytics industry named Simo Ahava.

I will summarize and interpret the information from his extensive post for you. Essentially, there are ways to log extra data into your google analytics that will be useful in adding depth and dimension that is unprecedented.

Custom Dimensions are fields that can exist in Google Analytics with values that you set, without requiring the API. Instead, we can use Google Analytics to run scripts and generate values from the information that is scraped off the cookie and page and inject it into the tracking code. These custom dimensions will then show up on your dashboard after a few days, and you can run reports on it.

The 4 Custom Dimensions that Simo refers to are:

  • Client ID
  • Session ID
  • Hit Timestamp
  • User ID

Client ID

The Client ID is based on the session metric that is a value already created by Google Analytics. Google Tag Manager can be set up to extract this value and added to the Custom Dimension. Gathering it does require custom JavaScript and loading triggers based on events… so it’s somewhat complicated. Unless you have it in a custom dimension, you cannot pull this data out of Google Analytics or filter it in any way since it is not exposed on the front end.

You would want to know this to get more detailed session information out of your users. Do not go through the trouble of getting this unless you plan on utilizing it. This leads us to the next Custom Dimension…

Session ID

The Session ID is also based on the session metric but is randomized in the way Simo is using it. The values are grouped into sessions, but they blend together in the front end. The only way to identify if two distinct visits belong to the same session is to ungroup these sessions. 

Hit Timestamp

Unlike the two Custom Dimensions before it, this dimension is hit-scoped, meaning it will be based on logged in visitors. Simo points out that this is done for privacy reasons so that I would stick with a hit-scoped method. The hit scope does require users to be logged in when they are visiting to get these metrics. You may be able to get this information by switching to a session or user-scoped view, but that is a “grey area,” as Simo says.

This value is found in Google Analytics but is not passed into the reporting. The value represents the exact moment that a tracking code is fired. You can use this to verify databases but also to see discrepancies between collected times and reported times. 

User ID

The final Custom Dimension is the User ID, from the Universal Analytics tracking that groups together the fires, sessions, and Client IDs. The reason you would add this is to tie everything you have done so far all together. The User ID matches a user to their session and timestamp by the Client ID. If you have a very complex network of web pages… these four methods allow you to merge them into a clearer picture. You should know that users are not the most accurate to begin with, which is why robust solutions are created to clean it up.

Step 4: Why phone calls as a sales generation method?

Limiting your sales generation to phone calls is a heavy constraint. However, there are some reasons you should consider generating phone calls as your main sales means from PPC campaigns:

  • You are an appointment based business.
  • You have a team that is trained to handle sales calls.
  • You have a team that is trained to handle sales calls.
  • You track your marketing phone calls and already have a budget dedicated to scale this.

Let’s talk about the last point, tracking your marketing phone calls because it’s something you can easily add-on to your business. Shameless plug aside, you can only improve your marketing by tracking it and having a plan in place ahead of time to utilize the findings.

There are a few other types of lead generation that can occur before you start making sales on the phone with them. Although we are assuming you have been tasked with generating hot leads for sales calls through PPC, you should also consider gathering leads through other means and then closing them on the phone.

Some of these basic methods employed include but are not limited to:

  • Form-fills to capture leads from landing pages
  • “Contact Us” phone calls or emails
  • Online chat platforms
  • SMS text messaging capturing

You don’t need to use all of these tactics, but consider how they would work for your company.

In Summary

If you must, I have outlined a detailed way to set up your website tracking for multiple sites of deployment using Google Tag Manager. Also, you’ve been shown how to add cross-domain tracking to your websites using that same tool. Finally, we did a deep dive into Simo Ahava’s solution for cleaning up the artifact-laden User information passed into Google Analytics using Custom Dimensions.

There are elements to help you improve your tracking so you have cleaner data that you can confidently say represents your actual web traffic. You can use this information to validate traffic counts no matter what business or model you deploy. If you have more tips, questions, or want to anything, you want to share? Go ahead and add it to the comments below.

Kevin Profile Shot
Kevin Dieny
Marketing Professional

Kevin is a contributing author and product expert in all things digital where he works as a Digital Marketing Analyst at CallSource.

How to Generate Hot Sales Calls from PPC Campaigns While Staying ROI Positive

How to Generate Hot Sales Calls from PPC Campaigns While Staying ROI Positive

Kevin Dieny

FIRM FIXED IDEAS

Pay Per Click
Lead Generation
Return on Investment

You have multiple priorities to keep in mind in marketing and sales – here’s some tips to keep both in the forefront of your mind.

“Generate hot sales calls from a PPC campaign while staying ROI positive.”

You are tasked with what seems impossible: “Generate hot sales calls from a PPC campaign while staying ROI positive.” If you worked at a marketing agency, then this would probably be in your everyday tasks while working with clients. But unless “Advertising,” “PPC,” “Traffic,” or “Affiliate” is in your job title, then I’m sure you have other things to focus on besides trying to generate calls from digital advertising.

Let’s jump into an extensive “how-to” for generating hot sales calls for those who have other things to focus on, as well as for experienced PPC managers looking to shore up what they know.

A task like “Generate hot sales calls” implies a current condition or state where this does not exist. “Scale hot sales calls” refers to a state where this is already occurring. I am going to assume that you are currently not generating hot sales calls from your PPC campaigns, so I’ll focus on the former. Since this is your task to work in an environment where something successful is desired but does not yet exist…welcome to marketing!

Step 1: Make a list of marketing assets before you start the project.

In marketing, you first have to start with an inventory of what you have to work with and go from there. Start with answering the following questions:

  • What is the budget for this task?
  • Who will be in charge of running the campaign and acquiring the creative assets?
  • What methods of digital advertising are we going to use? (Facebook, Google, LinkedIn, etc).
  • What marketing tools do we have access to assist us in this task? (Advertising tools, email marketing tools, landing page or thank you page creators, etc).
  • What is our sales process for handling the “hot sales calls” that we generate? What about the follow-up or missed opportunity sales calls?

After making this list, you might see some first steps that need to happen before you start generating hot phone call leads. You start with a list to identify the constraints you will be facing. The shorter your list in general, the more constraints you will find that crop up and delay or throw off your task. Have an answer for each bullet point above, and you will have enough to jump ahead.

Step 2: Defining a “Hot Sale” and their buyer intensity in the marketing funnel.

The more precise you define a “Hot Sale,” the better your PPC campaign will perform. We’ve previously talked about how to create better marketing qualified leads, but how do you ensure those MQLs become a “Hot Sale”? The quality of your leads is dependent on not only how well you can define what makes someone a qualified lead, but how well you can speak to these specific consumers with a relevant message.

We will use hot, warm, and cold to signify the buying intensity of sales prospects going forward.

Cold leads/prospects are at the top of the funnel and may not know your company, or there is a problem to fix.
Warm prospects need convincing and are most likely in the middle of the funnel.
Hot prospects are sales prospect further down the marketing funnel and ready to buy.

Let’s picture this funnel…

Each section of the funnel has a specific goal, and you might be surprised to find out that it is not always, “Call Me!”

Top of the funnel

The goal at the top of the funnel, where your cold prospects are located, is to make people AWARE. Awareness is defined as: informing people of a problem, of their current state, of a better state, and finally of who your company is and what you’re passionate about doing that nobody else does.

Middle of the funnel

The goal of the middle of the funnel is only to segment people into WARM or HOT groups. This is done by tracking the engagement types who you identify as someone who may be interested in buying. Deciding what to track is ultimately up to your company and what interactions consumers are having with your properties online. Draw a line in the sand and decide what is warm, or someone who needs more information, and hot, meaning someone who you confidently feel is ready to buy. Usually you won’t get this right the first time, but with trial and error and testing, you will get better….

Bottom of the funnel

The bottom of the funnel is where HOT audiences are sent. The goal at the bottom of the funnel is to send your prospect to a persuasive page to call you, set an appointment, or to engage with you in some fashion. The highly subjective part is being persuasive, but if you’ve done the work before this point and segmented them correctly, you should have sifted the leads with the highest buyer intensity.

Step 3: “Hot Leads” Consumer Messaging

“One Man’s Loss is Another Man’s Gain.” – Proverb

There are two categories you need to figure out to be successful in defining and speaking to your hottest consumers:

Relevance Value
Identify your target audience, discover their pain points, and speak to them in a way that solves for their needs quickly. Provide a quality offer to your target audience that will seem irresistible to them right now.

As referenced in the proverb above, your target audience should not be similar to anyone else’s. Even amongst your competition, your target audience should be distinct and aligned to what you sell.

Using the analogy of Subway and McDonalds, both are targeting more market share of the healthy fast food market—especially amongst the 18 to 30-year-old segment. Even still, one should be specifically going after the sandwich crowd and the other the burger and fries crowd. You see examples of this in commercials where advertising hones in on a specific segment of a targeted audience and tries to speak directly to them—such as the new “McVegan” burger that McDonald’s is testing or Subway’s “Fresh Fit” sandwich menu.

Although it is important to make your business stand out, don’t get too hung up on this step of really individualizing your messaging yet or blame this step later if your campaign doesn’t work out as intended. But I can’t give you this brief without also sharing a tool to help you make your marketing both relevant and valuable. I like to use DigitalMarketer’s Avatar Worksheet to stay on message and identify my hottest target audience.

Step 4: Why phone calls as a sales generation method?

Limiting your sales generation to phone calls is a heavy constraint. However, there are some reasons you should consider generating phone calls as your main sales means from PPC campaigns:

  • You are an appointment based business.
  • You have a team that is trained to handle sales calls.
  • You have a team that is trained to handle sales calls.
  • You track your marketing phone calls and already have a budget dedicated to scale this.

Let’s talk about the last point, tracking your marketing phone calls because it’s something you can easily add-on to your business. Shameless plug aside, you can only improve your marketing by tracking it and having a plan in place ahead of time to utilize the findings.

There are a few other types of lead generation that can occur before you start making sales on the phone with them. Although we are assuming you have been tasked with generating hot leads for sales calls through PPC, you should also consider gathering leads through other means and then closing them on the phone.

Some of these basic methods employed include but are not limited to:

  • Form-fills to capture leads from landing pages
  • “Contact Us” phone calls or emails
  • Online chat platforms
  • SMS text messaging capturing

You don’t need to use all of these tactics, but consider how they would work for your company.

Step 5: The Medium, Platforms, and Local of PPC Campaigns

We have now identified and explained all of the major constraints that a typical marketer would face in the task, “Generate hot sales calls from a PPC campaign.”

On to implementation…

My advice—plain and simple—is start with a local search segment of your targeted audience if possible. I recommend this from my experience in PPC. Don’t just take my word for it, here is a great article from ReviewTrackers on how local search is best for bottom of funnel.

Start with the easiest method to capture sales from PPC campaigns before getting more complex. Local advertisements are the easiest because you are overcoming a lot of the issues people have with PPC ads and their reluctance to trust them by saying you are a real company just around the corner from them.

Other platforms could work just as well for you and each of them has its pros and cons. Research by emarketer this last September details the growth of digital advertisement platforms and highlights a trend that digital ad spend is only increasing.

Taking it up a notch is the question of which platform would work best for generating hot sales calls?

I would recommend the following, depending on your target audiences:

  • Adwords for local search
  • Facebook for most everyone
  • LinkedIn for longer sales cycles or specific age groups
  • Bing for devices and age groups or even as a cheaper alternative to Adwords (not always cheaper)
  • Yelp for testing against Adwords
  • Instagram if you are a very visual company
  • SnapChat if you are gunning for millennials

…The point is go where your targeted audience is hiding, because your competitors probably are.

Finally, consider the medium you are utilizing for these PPC campaigns. I look at the content that is being linked to on multiple levels in order to test what works best for the structure of a keyword based search campaign:

  • The Brand Level
  • The Categorical Level
  • The Keyword Level

Brand Level
The brand is your company name and offers you the chance to target search marketing with a message to accurately portray your brand in a way that positions it for sales calls. Brand targeting should not be overlooked. When it comes to performance, they typically have the best Click Through Rates (CTR) and high-quality scores, which directly reduces your costs in running advertisements on those platforms. Someone who types in your company name into a search engine might be looking for the phone number… why not give them that opportunity and test this hypothesis?

Categorical Level
The category level should be made up of your products or topics that align with what your company does. If you do not have products that are clearly defined, consider looking at your mission statement for ideas. Assuming you have products, break them up into exclusive categories and make content and messaging to target an audience for each of those products.
Categories often coincide with price as lower priced products or services are easier entry points to a sale then a higher priced one. Targeting people for categories should take people to the product they searched for, and that corresponding landing page has to do a few things to be successful:

  • Make them aware of a problem they have.
  • Talk about the pain of that problem in the current state.
  • Refer to the ideal state (or the potential after).
  • Clearly make it your passion to help them solve this.
  • THEN… and only then… talk about your product as the solution you want to provide for a specific set of problems.

Keyword Level
The keyword level takes this even further, and so I do not recommend this for marketers just getting started. The reason that keyword level is more difficult is that you are isolating specific problems people are having through search terms and matches a highly relevant and valuable solution page for them.
People will type in hundreds of queries, so building quality content for all of even the most popular searches will take considerable time and effort. Every grouping of search queries has to be paired with a highly relevant piece of content which you have to build. For example:

Search Term: “personalized 833 vanity numbers”

Relevant Hypothetical Article: “Personalize your own 833 vanity number for your business [Article on Blog]”

Valuable Hypothetical Offer: “The Best 833 Vanity Numbers for Small Businesses [Downloadable List]”

In this example, the search term matched, a highly relevant article was displayed as an advertisement, the article makes mention of a downloadable offer for lead capture during and after the article’s content that sends people to a form to get their name, email, and phone number.

You would have to repeat this process for every valuable keyword in order to maximize the keyword level. These steps are simplified for categorical and brand as your need for content is much smaller since matched keywords are grouped into anything relevant to the category.

Before I move on from search marketing I do want to point out the click2call, call-only ads, and “call me now” buttons that are found in these platforms. My advice is to use these for remarketing lists for people lower down your funnel, but consider giving them the option to call you straight from the advertisement as well. Do not neglect the call extension, but also don’t jump straight to “call me!” unless you had a segment that was already hot.

Specificity is King

The more specific you get (Keyword > Category > Brand), the more personal you can make content and offers. You should be trying to speak to a specific, targeted group of people so that you get higher relevance and value.

What about the other platforms that are not search?

Image based advertisements
Image based advertisements require a landing page, and you should always do that! A landing page can be an article or a product detail page, but your consumers should be sent somewhere that they are likely to convert.

Retargeting
Retargeting is huge, and if you are not doing this, you are ignoring the ability to segment your visitors. I recommend implementing image ads as retargeting PPC campaigns because they really have to work hard. In comparison, search advertisements allow personalization by matching keywords that a consumer is looking for to a potential solution you provide.

In retargeting you can only segment that audience based on what the platform allows. Most platforms allow you to pixel an audience or upload a list that is matched to their database. By uploading or targeting specific pixeled visitors you can find people that are warm or hot based on their engagement.

I’ve seen clients go after their cancellations with some positive results because they have new pricing, a new product, or unique service addition that sets them apart. This means that as you conduct your business online, especially with advertisements, remember that marketing is a long term strategy and it often takes many touches to finally make a sale.

Adding touches does not always guarantee a Hot lead or a sale but it does improve the chances that marketing will be successful. Your purpose for this whole PPC campaign task is to build a marketing strategy that will inform you enough to know how many touches on average it takes for someone to become your next client. You can only know this if you track your marketing activities.

Step 6: What does ROI positive mean for marketing?

Marketing is a long-term strategy!!!

Okay, I get it, but I need results tomorrow, so what should I do?

A key element I have purposely left out of all these steps is the time frame. I have left it out because technically you could be ROI positive if you take a long time, with highly calculated steps, and spent just enough at each stage to get what you want. That oversimplified explanation can last months or years as marketing teams make incremental changes until they achieve return on investments (ROI) that are double, quadruple, or even 100x their spend.

That’s impossible… 100x?

The truth is products that take double or triple-digit returns are usually able to sell themselves. I wouldn’t always credit a marketing team for that kind of return unless they started out with utterly abysmal marketing and did a 180.

Achieving ROI is based primarily on the LTV (lifetime value of a customer). Once you know how much value an average sale is bringing you, you have your breakeven point for ad spend. You need to break-up your budget accordingly for testing.

The continuum that marketing lives on is that as time is shortened costs usually go up. If you want to be ROI positive tomorrow, your chances are very slim.

My advice is to consider time in your equation.

In order to become ROI positive and start multiplying your return, you need a well-oiled marketing machine. I have laid out the steps that you need to master, and I mean it when I say you must master them. You have to assume that you will fail along the way – the key is failing quick. Learn from your mistakes, take calculated risks, and plan on increasing your budgets only after you have made wins.

Marketing optimization is a job all its own.

Most people think marketing ends when an advertisement is first turned on and calls start coming in (or maybe they don’t); this is when optimization starts. You will only improve what you bother to track. Here are recommendations for constant improvement and to get ROI positive going forward:

  • A/B Testing (avoid multivariate until you have a large team or lots of content assets)
  • Content Testing
  • Time of Day Testing
  • Day of the Week Testing
  • Format Testing (videos, infographics, television ads, etc.)
  • Cannon Ball Testing

Cannon ball testing is when you make a dramatic shift (as opposed to iterative testing) and change a ton of things all at once. Analysts scream and optimization specialists jump out of the window when they hear you are doing this… but sometimes you have to. You need to experiment with changes and sometimes you have to make big shifts until you find some semblance of success.

Summarize everything for me for watercooler conversation.

In summary, put on your lucky thinking cap and go down this step by step list to learn how to generate hot sales calls from PPC campaigns while staying ROI positive:

  • Make a list of marketing assets before you start the project
  • Defining a “Hot Sale” and their buyer intensity in the funnel
  • “Hot Leads” Consumer Messaging
  • Why phone calls as a sales generation method?
  • The Medium, Platforms, and Local of PPC Campaigns (The MVP)
  • What does ROI positive mean for marketers?

By the end of this when someone says, “What is your return on ad spend (ROAS)?” or “What is your return on investment (ROI)?” you can confidently answer and have the insight to explain how you are actively improving. Even if it’s not rainbows or pots of gold today, you can celebrate the fact that you are making attempts and failing fast because marketing is a long-term strategy and you will get there!

Kevin Profile Shot
Kevin Dieny
Marketing Professional

Kevin is a contributing author and product expert in all things digital where he works as a Digital Marketing Analyst at CallSource.

Software vs. Human Call Processing: What’s the Difference?

Software vs. Human Call Processing: What’s the Difference?

Kevin Dieny

FIRM FIXED IDEAS

Speech to text
Speech analytics
Software call processing

Software and human call processing have significant differences for businesses who rely on the accuracy of phone call information.

“With the introduction of Apple’s Siri and similar voice search services from Google and Microsoft, it is natural to wonder why it has taken so long for voice recognition technology to advance to this level, and we wonder, when can we expect to hear a more human-level performance?” – Baker, Huang, and Reddy

The difference between software speech analytics and human call processing is wide, but there are improvements. Technology has come a long way in the past 40 years making enormous leaps in speech analytics. Companies that find the newest system are quick to patent and secure their advances as the demand for speech-enabled devices grows year after year. Mobile devices are increasingly adding advanced speech analytics to enhance productivity, make driving safer, and texting hands-free.

The goal of speech analytics for businesses is to affordably identify what happened on a phone call, if the caller a missed opportunity, and what this information can do to help both marketing and sales closing going forward. The goal of the patent creators is to be a little more accurate and have fewer errors than their competitors to have a superior product. These goals clash with the largest innovators of speech analytics technology whose goal is to make speech analytics better than human processing.

“Speech analytics in the next 40 years will pass the Turning test.” – Baker, Huang, and Reddy

There are three major problems that software speech analytics has to overcome: background noise, echo or reverberation, and the accent or dialect variations. Major scientific theories, algorithms, and models have taken shape around advances in modern computing allowing innovative ideas to finally become a reality. In the following sections, we will discuss these three major problems that you should consider if you are interested in call processing.

“The basic learning and decoding algorithms have not changed substantially in 40 years.” – Baker, Huang, and Reddy

To properly discuss how the differences impact businesses who utilize speech analytics to score and process phone calls we need to know their WER. The Word Error Rate (WER) is a standardized model of assessing how well software performs at speech analytics. “The word error rate (WER) is a good measure of the performance of dictation system, and a reasonable approximation of the success of a voice search system.” – Senior

“The best commercial speech analytics systems achieve 30.5% error.” – Case

Background Noise: How well do Apple, Google, and Microsoft perform?

Background noise is a contributing factor to the ability to clearly and accurately transcribe a conversation during a phone call. Sources of noise include wind, crowds, music, and even screaming children. Ideally, a phone call is placed by the caller in a quiet place where they can think and talk coherently.

That’s not always the case.

Looking above at figure 1, the ability for software speech analytics to accurately translate what is being said by the caller in noisy environments corresponds to high WER, or high error rates. When the caller knows that they have difficulty being heard by the person on the other side of the line, they will attempt to compensate, and this is called the Lombard effect. The Lombard effect often makes software recognition even more difficult because the caller’s speech fluctuates:

  • The loudness of the caller’s voice goes up and down.
  • Pitch changes in the caller’s voice.
  • The harmonic rate of words changes for the caller.
  • The duration and pausing of syllable intensity shifts for the caller.

The big takeaway of whether or not noise is a consideration for you and your business is how often has noise impacted your companies calls in the past, and what are you going to do about it?

“Results of our study shows that performance of cloud-based speech analytics systems can be affected by jitter and packet loss; which are commonly occurring over WiFi and cellular and mobile network connections.” – Assefi

If even 25% of your businesses phone calls occur from cell phones, your business has been negatively impacted by noise in some degree – but it is ultimately up to you to put a dollar value on that expense. Having human call processing analysts screen and listen to your calls is one way to mitigate the factor of noise in calls.

Echo and Reverberation: Building Robust Call Handling Systems

Software speech analytics software must account for the direction of the voice. A simple way to understand this problem is to carry on a conversation in an empty and hardwood floored house. If you start to hear your voice echo off the walls and floor, it interrupts what you are saying. In the image above, the Airforce tests echo and reverberation in an echo-free room. As you can see, people typically won’t be calling your business from this type of place.

Imagine how difficult it would be to hear someone’s voice with an echo also being picked up by the phone microphone and relayed over the call. Hello… hello… hello…

The best takeaway from echo and reverberation I can give you is that to train software on how to account for the doubling effect of sounds from echo, recordings of human scored calls are used over and over again, every single day to add to training data. As of today, software is 100% dependent upon calls that were already scored by humans to raise their day-to-day accuracy of phone call transcriptions using these steps:

  1. A human call analyst is used to evaluate how well a system is at diagnosing echo and sound reverberation in the caller’s environment.
  2. After the phone call has been scored and transcribed by a human call analyst, the information is added to software speech analytics training data.
  3. The speech software will then attempt to transcribe the call and differences between the human transcription and the software transcription produces a WER rate.
  4. Rinse and repeat… over and over again millions of times to lower the average WER rate.

Human call analysts are trained and vetted using a similar school of thought using these steps:

  1. Human call analysts vetted with years of call handling experience are put through a school with hours of training with typical calls, difficult calls, and how to avoid word errors.
  2. CallSource human call analysts are given amazing opportunities to work from their chosen locations, they are screened and trained on the best practices in the industry and are motivated to improve their accuracy rates continually.
  3. Once human call analysts are certified, they begin taking phone calls and work with mentors and colleagues to ensure that processes are being followed.
  4. Rinse and repeat… on their schedule, delivering the lowest WER rates in the industry.

“The network must not only learn how to recognize speech sounds, but how to transform them into letters, this is challenging, especially in an orthographically irregular language like English.” – Alex Graves

Graves (quoted above) has researched for improving models of speech analytics with neural networks and gives insight into how speech analytics output can defer in decoding single sentences:

Example #1

Speaker: TO ILLUSTRATE THE POINT A PROMINENT MIDDLE EAST ANALYST IN WASHINGTON RECEIVES A CALL FROM ONE CAMPAIGN.

Speech Recognition Software: TWO ALSTRAIT THE POINT A PROMINENT MIDILLE EAST ANALYST IM WASHINGTON RECOUNCACALL FROM ONE CAMPAIGN.

 

Example #2

Speaker: ALL THE EQUITY RAISING IN MILAN GAVE THAT STOCK MARKET INDIGESTION LAST YEAR.

Speech Recognition Software: ALL THE EQUITY RAISING IN MULONG GAVE THAT STACKR MARKET IN JUSTIAN LAST YEAR.

Source: Towards End-to-End Speech Recognition with Recurrent Neural Networks by Alex Graves

Both software speech analytics and human call analysts require robust systems to ensure that WER rates are as low as possible. WER rates directly contribute to businesses missing phone calls and misunderstanding what actually happened on phone calls. End to end speech software struggles to deal with echo and state of the art solutions continue to fall short the worse the noise and echo are in the environment of the caller.

Accents and Dialects: The curveball of call handling

According to Wikipedia, the United States has over 30 major dialects of the English language. For native-born Americans, these apply to the geographic location you are born and raised, but also to the dialects of your parents. Assuming that all of your business calls are from callers with a dialect that software has trained on, you should be looking a moderately high but acceptable WER rate for accents. The diversity of your callers weighs heavily on the accuracy outcomes of speech analytics WER rates. You would see a shocking rise in WER rates for callers that come from a distinct dialect or who carry a unique accent separate from what the software has been training on.

“The performance of speech analytics systems degrades when speaker accent is different from that in the training set. Accent-independent or accent-dependent recognition both require collection of more training data.” – Kiu Wai Kat

Accents and dialects represent a curve, such as a curveball, as utterances are spoken, and software attempts to accurately decode the words into coherent sentences. The degrading accuracy rates contribute to large gaps and word errors which may completely miss what was said or intended. When the outcome of what happened is dependent on a single word… and that word is usually interpreted incorrectly; the consequences are detrimental for businesses.

All of the large studies on accents point out that speech analytics has been unable to conquer the British English speech systems of Scotland, and with hilarious results (see the video below, video contains adult humor)

 

Only small progress has been made in dealing with different dialects and accents as these interrupt the way that words sound to software speech analytics. The systematic approach required for robust software speech analytics is challenged by the need for a system that is adaptable to a large variance of pronunciation.

Human call analysts also struggle with accents and dialects unfamiliar to them. However, one huge advantage that CallSource has is to select and hire human analysts who are familiar with those accents. There are call analysts who understand large sets of accents and dialects with ease and they are fully capable of the challenge.

“The error rate from accent speakers is around 30.89%” – Liu Wai Kat

The quantifiable reasons that accents are difficult to classify, transcribe and decode have to do with the acoustic differences between accent groups. Those differences are difficult to account for while still accounting for noise and echo in the environment. For ESL (English as a second language) speakers who have thicker accents, the problems are compounded if they are not in ideally quiet and low echo environments.

The detection of key phrases during a call is vitally important to understanding what happened during a call. How those key phrases are used is how that call will ultimately be classified by the system. Marketing and sales are unable to move forward effectively without accurate information of what happened during an initial phone call into the business. Advances in the improvement of accents will have to be developed in every accent and dialect individually and added to training data collectively to overcome the problem of high WER rates.

Conclusion

You should ask yourself: are you clever enough to handle high phone call accuracy rates? Can you make a difference in your business by knowing what happened on every single call? Can you achieve the results you need even with pages of word errors?

We think you are clever.

We know that you could turn that knowledge into practical business decisions for the future, and by those decisions, you can make waves in your market. Assuming you do not want to deal with all of these issues, just go back to doing what you do best:

  • Asking the caller to call you back from a quieter place over 30% of the time.
  • Asking the caller to move into a room with fewer echoes over 30% of the time.
  • Having to task a sales or service employee to re-listen to 30% of phone calls from a region with specific accents.
  • Asking the caller to call you back from a landline or a location with better reception.

The differences between software-based speech analytics and human call analysts comes down to how much of an impact accuracy makes in your business. Talk to your appointment setters and ask them if they ever have trouble hearing what people are saying because it’s a good bet that if they have ever had trouble, your call processing has had trouble.

Article References:

  1. Assefi, Mehdi, et al. “An Experimental Evaluation of Apple Siri and Google Speech analytics.” An Experimental Evaluation of Apple Siri and Google Speech analytics, www.cs.montana.edu/izurieta/pubs/sede2_2015.pdf.
  2. Hannun, Awni, et al. “Deep Speech: Scaling up End-to-End Speech analytics.” [1412.5567] Deep Speech: Scaling up End-to-End Speech analytics, 19 Dec. 2014, arxiv.org/abs/1412.5567.
  3. Kat, Liu Wai, and P. Fung. “Fast Accent Identification and Accented Speech analytics.” 1999 IEEE International Conference on Acoustics, Speech, and Signal Processing. Proceedings. ICASSP99 (Cat. No.99CH36258), 1999, doi:10.1109/icassp.1999.758102.
  4. Senior, Andrew, et al. “An Empirical Study of Learning Rates in Deep Neural Networks for Speech analytics.” An Empirical Study of Learning Rates in Deep Neural Networks for Speech analytics – IEEE Conference Publication, 2013 IEEE International Conference on Acoustics, Speech and Signal Processing, ieeexplore.ieee.org/document/6638963/.
  5. Xuedong Huang, James Baker, and Raj Reddy. 2014. A historical perspective of speech analytics. Commun. ACM 57, 1 (January 2014), 94-103. DOI: https://doi.org/10.1145/2500887
  6. Zhang, Ying, et al. “Towards End-to-End Speech analytics with Deep Convolutional Neural Networks.” [1701.02720] Towards End-to-End Speech analytics with Deep Convolutional Neural Networks, Cornell University, 10 Jan. 2017, arxiv.org/abs/1701.02720.
Kevin Profile Shot
Kevin Dieny
Marketing Professional

Kevin is a contributing author and product expert in all things digital where he works as a Digital Marketing Analyst at CallSource.